-
Notifications
You must be signed in to change notification settings - Fork 196
/
punctuator.py
176 lines (124 loc) · 5.31 KB
/
punctuator.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
# coding: utf-8
from __future__ import division, print_function
import models
import data
import theano
import sys
from io import open
import theano.tensor as T
import numpy as np
MAX_SUBSEQUENCE_LEN = 200
def to_array(arr, dtype=np.int32):
# minibatch of 1 sequence as column
return np.array([arr], dtype=dtype).T
def convert_punctuation_to_readable(punct_token):
if punct_token == data.SPACE:
return " "
else:
return punct_token[0]
def restore_with_pauses(output_file, text, pauses, word_vocabulary, reverse_punctuation_vocabulary, predict_function):
i = 0
with open(output_file, 'w', encoding='utf-8') as f_out:
while True:
subsequence = text[i:i+MAX_SUBSEQUENCE_LEN]
subsequence_pauses = pauses[i:i+MAX_SUBSEQUENCE_LEN]
if len(subsequence) == 0:
break
converted_subsequence = [word_vocabulary.get(w, word_vocabulary[data.UNK]) for w in subsequence]
y = predict_function(to_array(converted_subsequence), to_array(subsequence_pauses, dtype=theano.config.floatX))
f_out.write(subsequence[0])
last_eos_idx = 0
punctuations = []
for y_t in y:
p_i = np.argmax(y_t.flatten())
punctuation = reverse_punctuation_vocabulary[p_i]
punctuations.append(punctuation)
if punctuation in data.EOS_TOKENS:
last_eos_idx = len(punctuations) # we intentionally want the index of next element
if subsequence[-1] == data.END:
step = len(subsequence) - 1
elif last_eos_idx != 0:
step = last_eos_idx
else:
step = len(subsequence) - 1
for j in range(step):
f_out.write(" " + punctuations[j] + " " if punctuations[j] != data.SPACE else " ")
if j < step - 1:
f_out.write(subsequence[1+j])
if subsequence[-1] == data.END:
break
i += step
def restore(output_file, text, word_vocabulary, reverse_punctuation_vocabulary, predict_function):
i = 0
with open(output_file, 'w', encoding='utf-8') as f_out:
while True:
subsequence = text[i:i+MAX_SUBSEQUENCE_LEN]
if len(subsequence) == 0:
break
converted_subsequence = [word_vocabulary.get(w, word_vocabulary[data.UNK]) for w in subsequence]
y = predict_function(to_array(converted_subsequence))
f_out.write(subsequence[0])
last_eos_idx = 0
punctuations = []
for y_t in y:
p_i = np.argmax(y_t.flatten())
punctuation = reverse_punctuation_vocabulary[p_i]
punctuations.append(punctuation)
if punctuation in data.EOS_TOKENS:
last_eos_idx = len(punctuations) # we intentionally want the index of next element
if subsequence[-1] == data.END:
step = len(subsequence) - 1
elif last_eos_idx != 0:
step = last_eos_idx
else:
step = len(subsequence) - 1
for j in range(step):
f_out.write(" " + punctuations[j] + " " if punctuations[j] != data.SPACE else " ")
if j < step - 1:
f_out.write(subsequence[1+j])
if subsequence[-1] == data.END:
break
i += step
if __name__ == "__main__":
if len(sys.argv) > 1:
model_file = sys.argv[1]
else:
sys.exit("Model file path argument missing")
if len(sys.argv) > 2:
output_file = sys.argv[2]
else:
sys.exit("Output file path argument missing")
use_pauses = len(sys.argv) > 3 and bool(int(sys.argv[3]))
x = T.imatrix('x')
if use_pauses:
p = T.matrix('p')
print("Loading model parameters...")
net, _ = models.load(model_file, 1, x, p)
print("Building model...")
predict = theano.function(
inputs=[x, p],
outputs=net.y
)
else:
print("Loading model parameters...")
net, _ = models.load(model_file, 1, x)
print("Building model...")
predict = theano.function(
inputs=[x],
outputs=net.y
)
word_vocabulary = net.x_vocabulary
punctuation_vocabulary = net.y_vocabulary
reverse_word_vocabulary = {v:k for k,v in word_vocabulary.items()}
reverse_punctuation_vocabulary = {v:k for k,v in punctuation_vocabulary.items()}
input_text = open(sys.stdin.fileno(), 'r', encoding='utf-8').read()
if len(input_text) == 0:
sys.exit("Input text from stdin missing.")
text = [w for w in input_text.split() if w not in punctuation_vocabulary and w not in data.PUNCTUATION_MAPPING and not w.startswith(data.PAUSE_PREFIX)] + [data.END]
pauses = [float(s.replace(data.PAUSE_PREFIX,"").replace(">","")) for s in input_text.split() if s.startswith(data.PAUSE_PREFIX)]
if not use_pauses:
restore(output_file, text, word_vocabulary, reverse_punctuation_vocabulary, predict)
else:
if not pauses:
pauses = [0.0 for _ in range(len(text)-1)]
restore_with_pauses(output_file, text, pauses, word_vocabulary, reverse_punctuation_vocabulary, predict)